xref: /linux-6.15/include/linux/ceph/rados.h (revision bb7e5ce7)
1 #ifndef CEPH_RADOS_H
2 #define CEPH_RADOS_H
3 
4 /*
5  * Data types for the Ceph distributed object storage layer RADOS
6  * (Reliable Autonomic Distributed Object Store).
7  */
8 
9 #include <linux/ceph/msgr.h>
10 
11 /*
12  * fs id
13  */
14 struct ceph_fsid {
15 	unsigned char fsid[16];
16 };
17 
18 static inline int ceph_fsid_compare(const struct ceph_fsid *a,
19 				    const struct ceph_fsid *b)
20 {
21 	return memcmp(a, b, sizeof(*a));
22 }
23 
24 /*
25  * ino, object, etc.
26  */
27 typedef __le64 ceph_snapid_t;
28 #define CEPH_SNAPDIR ((__u64)(-1))  /* reserved for hidden .snap dir */
29 #define CEPH_NOSNAP  ((__u64)(-2))  /* "head", "live" revision */
30 #define CEPH_MAXSNAP ((__u64)(-3))  /* largest valid snapid */
31 
32 struct ceph_timespec {
33 	__le32 tv_sec;
34 	__le32 tv_nsec;
35 } __attribute__ ((packed));
36 
37 
38 /*
39  * object layout - how objects are mapped into PGs
40  */
41 #define CEPH_OBJECT_LAYOUT_HASH     1
42 #define CEPH_OBJECT_LAYOUT_LINEAR   2
43 #define CEPH_OBJECT_LAYOUT_HASHINO  3
44 
45 /*
46  * pg layout -- how PGs are mapped onto (sets of) OSDs
47  */
48 #define CEPH_PG_LAYOUT_CRUSH  0
49 #define CEPH_PG_LAYOUT_HASH   1
50 #define CEPH_PG_LAYOUT_LINEAR 2
51 #define CEPH_PG_LAYOUT_HYBRID 3
52 
53 #define CEPH_PG_MAX_SIZE      32  /* max # osds in a single pg */
54 
55 /*
56  * placement group.
57  * we encode this into one __le64.
58  */
59 struct ceph_pg_v1 {
60 	__le16 preferred; /* preferred primary osd */
61 	__le16 ps;        /* placement seed */
62 	__le32 pool;      /* object pool */
63 } __attribute__ ((packed));
64 
65 /*
66  * pg_pool is a set of pgs storing a pool of objects
67  *
68  *  pg_num -- base number of pseudorandomly placed pgs
69  *
70  *  pgp_num -- effective number when calculating pg placement.  this
71  * is used for pg_num increases.  new pgs result in data being "split"
72  * into new pgs.  for this to proceed smoothly, new pgs are intiially
73  * colocated with their parents; that is, pgp_num doesn't increase
74  * until the new pgs have successfully split.  only _then_ are the new
75  * pgs placed independently.
76  *
77  *  lpg_num -- localized pg count (per device).  replicas are randomly
78  * selected.
79  *
80  *  lpgp_num -- as above.
81  */
82 #define CEPH_NOPOOL  ((__u64) (-1))  /* pool id not defined */
83 
84 #define CEPH_POOL_TYPE_REP     1
85 #define CEPH_POOL_TYPE_RAID4   2 /* never implemented */
86 #define CEPH_POOL_TYPE_EC      3
87 
88 /*
89  * stable_mod func is used to control number of placement groups.
90  * similar to straight-up modulo, but produces a stable mapping as b
91  * increases over time.  b is the number of bins, and bmask is the
92  * containing power of 2 minus 1.
93  *
94  * b <= bmask and bmask=(2**n)-1
95  * e.g., b=12 -> bmask=15, b=123 -> bmask=127
96  */
97 static inline int ceph_stable_mod(int x, int b, int bmask)
98 {
99 	if ((x & bmask) < b)
100 		return x & bmask;
101 	else
102 		return x & (bmask >> 1);
103 }
104 
105 /*
106  * object layout - how a given object should be stored.
107  */
108 struct ceph_object_layout {
109 	struct ceph_pg_v1 ol_pgid;   /* raw pg, with _full_ ps precision. */
110 	__le32 ol_stripe_unit;    /* for per-object parity, if any */
111 } __attribute__ ((packed));
112 
113 /*
114  * compound epoch+version, used by storage layer to serialize mutations
115  */
116 struct ceph_eversion {
117 	__le64 version;
118 	__le32 epoch;
119 } __attribute__ ((packed));
120 
121 /*
122  * osd map bits
123  */
124 
125 /* status bits */
126 #define CEPH_OSD_EXISTS  (1<<0)
127 #define CEPH_OSD_UP      (1<<1)
128 #define CEPH_OSD_AUTOOUT (1<<2)  /* osd was automatically marked out */
129 #define CEPH_OSD_NEW     (1<<3)  /* osd is new, never marked in */
130 
131 extern const char *ceph_osd_state_name(int s);
132 
133 /* osd weights.  fixed point value: 0x10000 == 1.0 ("in"), 0 == "out" */
134 #define CEPH_OSD_IN  0x10000
135 #define CEPH_OSD_OUT 0
136 
137 /* osd primary-affinity.  fixed point value: 0x10000 == baseline */
138 #define CEPH_OSD_MAX_PRIMARY_AFFINITY 0x10000
139 #define CEPH_OSD_DEFAULT_PRIMARY_AFFINITY 0x10000
140 
141 
142 /*
143  * osd map flag bits
144  */
145 #define CEPH_OSDMAP_NEARFULL (1<<0)  /* sync writes (near ENOSPC) */
146 #define CEPH_OSDMAP_FULL     (1<<1)  /* no data writes (ENOSPC) */
147 #define CEPH_OSDMAP_PAUSERD  (1<<2)  /* pause all reads */
148 #define CEPH_OSDMAP_PAUSEWR  (1<<3)  /* pause all writes */
149 #define CEPH_OSDMAP_PAUSEREC (1<<4)  /* pause recovery */
150 #define CEPH_OSDMAP_NOUP     (1<<5)  /* block osd boot */
151 #define CEPH_OSDMAP_NODOWN   (1<<6)  /* block osd mark-down/failure */
152 #define CEPH_OSDMAP_NOOUT    (1<<7)  /* block osd auto mark-out */
153 #define CEPH_OSDMAP_NOIN     (1<<8)  /* block osd auto mark-in */
154 #define CEPH_OSDMAP_NOBACKFILL (1<<9) /* block osd backfill */
155 #define CEPH_OSDMAP_NORECOVER (1<<10) /* block osd recovery and backfill */
156 #define CEPH_OSDMAP_NOSCRUB  (1<<11) /* block periodic scrub */
157 #define CEPH_OSDMAP_NODEEP_SCRUB (1<<12) /* block periodic deep-scrub */
158 #define CEPH_OSDMAP_NOTIERAGENT (1<<13) /* disable tiering agent */
159 #define CEPH_OSDMAP_NOREBALANCE (1<<14) /* block osd backfill unless pg is degraded */
160 #define CEPH_OSDMAP_SORTBITWISE (1<<15) /* use bitwise hobject_t sort */
161 #define CEPH_OSDMAP_REQUIRE_JEWEL    (1<<16) /* require jewel for booting osds */
162 #define CEPH_OSDMAP_REQUIRE_KRAKEN   (1<<17) /* require kraken for booting osds */
163 #define CEPH_OSDMAP_REQUIRE_LUMINOUS (1<<18) /* require l for booting osds */
164 #define CEPH_OSDMAP_RECOVERY_DELETES (1<<19) /* deletes performed during recovery instead of peering */
165 
166 /*
167  * The error code to return when an OSD can't handle a write
168  * because it is too large.
169  */
170 #define OSD_WRITETOOBIG EMSGSIZE
171 
172 /*
173  * osd ops
174  *
175  * WARNING: do not use these op codes directly.  Use the helpers
176  * defined below instead.  In certain cases, op code behavior was
177  * redefined, resulting in special-cases in the helpers.
178  */
179 #define CEPH_OSD_OP_MODE       0xf000
180 #define CEPH_OSD_OP_MODE_RD    0x1000
181 #define CEPH_OSD_OP_MODE_WR    0x2000
182 #define CEPH_OSD_OP_MODE_RMW   0x3000
183 #define CEPH_OSD_OP_MODE_SUB   0x4000
184 #define CEPH_OSD_OP_MODE_CACHE 0x8000
185 
186 #define CEPH_OSD_OP_TYPE       0x0f00
187 #define CEPH_OSD_OP_TYPE_LOCK  0x0100
188 #define CEPH_OSD_OP_TYPE_DATA  0x0200
189 #define CEPH_OSD_OP_TYPE_ATTR  0x0300
190 #define CEPH_OSD_OP_TYPE_EXEC  0x0400
191 #define CEPH_OSD_OP_TYPE_PG    0x0500
192 #define CEPH_OSD_OP_TYPE_MULTI 0x0600 /* multiobject */
193 
194 #define __CEPH_OSD_OP1(mode, nr) \
195 	(CEPH_OSD_OP_MODE_##mode | (nr))
196 
197 #define __CEPH_OSD_OP(mode, type, nr) \
198 	(CEPH_OSD_OP_MODE_##mode | CEPH_OSD_OP_TYPE_##type | (nr))
199 
200 #define __CEPH_FORALL_OSD_OPS(f)					    \
201 	/** data **/							    \
202 	/* read */							    \
203 	f(READ,		__CEPH_OSD_OP(RD, DATA, 1),	"read")		    \
204 	f(STAT,		__CEPH_OSD_OP(RD, DATA, 2),	"stat")		    \
205 	f(MAPEXT,	__CEPH_OSD_OP(RD, DATA, 3),	"mapext")	    \
206 									    \
207 	/* fancy read */						    \
208 	f(MASKTRUNC,	__CEPH_OSD_OP(RD, DATA, 4),	"masktrunc")	    \
209 	f(SPARSE_READ,	__CEPH_OSD_OP(RD, DATA, 5),	"sparse-read")	    \
210 									    \
211 	f(NOTIFY,	__CEPH_OSD_OP(RD, DATA, 6),	"notify")	    \
212 	f(NOTIFY_ACK,	__CEPH_OSD_OP(RD, DATA, 7),	"notify-ack")	    \
213 									    \
214 	/* versioning */						    \
215 	f(ASSERT_VER,	__CEPH_OSD_OP(RD, DATA, 8),	"assert-version")   \
216 									    \
217 	f(LIST_WATCHERS, __CEPH_OSD_OP(RD, DATA, 9),	"list-watchers")    \
218 									    \
219 	f(LIST_SNAPS,	__CEPH_OSD_OP(RD, DATA, 10),	"list-snaps")	    \
220 									    \
221 	/* sync */							    \
222 	f(SYNC_READ,	__CEPH_OSD_OP(RD, DATA, 11),	"sync_read")	    \
223 									    \
224 	/* write */							    \
225 	f(WRITE,	__CEPH_OSD_OP(WR, DATA, 1),	"write")	    \
226 	f(WRITEFULL,	__CEPH_OSD_OP(WR, DATA, 2),	"writefull")	    \
227 	f(TRUNCATE,	__CEPH_OSD_OP(WR, DATA, 3),	"truncate")	    \
228 	f(ZERO,		__CEPH_OSD_OP(WR, DATA, 4),	"zero")		    \
229 	f(DELETE,	__CEPH_OSD_OP(WR, DATA, 5),	"delete")	    \
230 									    \
231 	/* fancy write */						    \
232 	f(APPEND,	__CEPH_OSD_OP(WR, DATA, 6),	"append")	    \
233 	f(SETTRUNC,	__CEPH_OSD_OP(WR, DATA, 8),	"settrunc")	    \
234 	f(TRIMTRUNC,	__CEPH_OSD_OP(WR, DATA, 9),	"trimtrunc")	    \
235 									    \
236 	f(TMAPUP,	__CEPH_OSD_OP(RMW, DATA, 10),	"tmapup")	    \
237 	f(TMAPPUT,	__CEPH_OSD_OP(WR, DATA, 11),	"tmapput")	    \
238 	f(TMAPGET,	__CEPH_OSD_OP(RD, DATA, 12),	"tmapget")	    \
239 									    \
240 	f(CREATE,	__CEPH_OSD_OP(WR, DATA, 13),	"create")	    \
241 	f(ROLLBACK,	__CEPH_OSD_OP(WR, DATA, 14),	"rollback")	    \
242 									    \
243 	f(WATCH,	__CEPH_OSD_OP(WR, DATA, 15),	"watch")	    \
244 									    \
245 	/* omap */							    \
246 	f(OMAPGETKEYS,	__CEPH_OSD_OP(RD, DATA, 17),	"omap-get-keys")    \
247 	f(OMAPGETVALS,	__CEPH_OSD_OP(RD, DATA, 18),	"omap-get-vals")    \
248 	f(OMAPGETHEADER, __CEPH_OSD_OP(RD, DATA, 19),	"omap-get-header")  \
249 	f(OMAPGETVALSBYKEYS, __CEPH_OSD_OP(RD, DATA, 20), "omap-get-vals-by-keys") \
250 	f(OMAPSETVALS,	__CEPH_OSD_OP(WR, DATA, 21),	"omap-set-vals")    \
251 	f(OMAPSETHEADER, __CEPH_OSD_OP(WR, DATA, 22),	"omap-set-header")  \
252 	f(OMAPCLEAR,	__CEPH_OSD_OP(WR, DATA, 23),	"omap-clear")	    \
253 	f(OMAPRMKEYS,	__CEPH_OSD_OP(WR, DATA, 24),	"omap-rm-keys")	    \
254 	f(OMAP_CMP,	__CEPH_OSD_OP(RD, DATA, 25),	"omap-cmp")	    \
255 									    \
256 	/* tiering */							    \
257 	f(COPY_FROM,	__CEPH_OSD_OP(WR, DATA, 26),	"copy-from")	    \
258 	f(COPY_GET_CLASSIC, __CEPH_OSD_OP(RD, DATA, 27), "copy-get-classic") \
259 	f(UNDIRTY,	__CEPH_OSD_OP(WR, DATA, 28),	"undirty")	    \
260 	f(ISDIRTY,	__CEPH_OSD_OP(RD, DATA, 29),	"isdirty")	    \
261 	f(COPY_GET,	__CEPH_OSD_OP(RD, DATA, 30),	"copy-get")	    \
262 	f(CACHE_FLUSH,	__CEPH_OSD_OP(CACHE, DATA, 31),	"cache-flush")	    \
263 	f(CACHE_EVICT,	__CEPH_OSD_OP(CACHE, DATA, 32),	"cache-evict")	    \
264 	f(CACHE_TRY_FLUSH, __CEPH_OSD_OP(CACHE, DATA, 33), "cache-try-flush") \
265 									    \
266 	/* convert tmap to omap */					    \
267 	f(TMAP2OMAP,	__CEPH_OSD_OP(RMW, DATA, 34),	"tmap2omap")	    \
268 									    \
269 	/* hints */							    \
270 	f(SETALLOCHINT,	__CEPH_OSD_OP(WR, DATA, 35),	"set-alloc-hint")   \
271 									    \
272 	/** multi **/							    \
273 	f(CLONERANGE,	__CEPH_OSD_OP(WR, MULTI, 1),	"clonerange")	    \
274 	f(ASSERT_SRC_VERSION, __CEPH_OSD_OP(RD, MULTI, 2), "assert-src-version") \
275 	f(SRC_CMPXATTR,	__CEPH_OSD_OP(RD, MULTI, 3),	"src-cmpxattr")	    \
276 									    \
277 	/** attrs **/							    \
278 	/* read */							    \
279 	f(GETXATTR,	__CEPH_OSD_OP(RD, ATTR, 1),	"getxattr")	    \
280 	f(GETXATTRS,	__CEPH_OSD_OP(RD, ATTR, 2),	"getxattrs")	    \
281 	f(CMPXATTR,	__CEPH_OSD_OP(RD, ATTR, 3),	"cmpxattr")	    \
282 									    \
283 	/* write */							    \
284 	f(SETXATTR,	__CEPH_OSD_OP(WR, ATTR, 1),	"setxattr")	    \
285 	f(SETXATTRS,	__CEPH_OSD_OP(WR, ATTR, 2),	"setxattrs")	    \
286 	f(RESETXATTRS,	__CEPH_OSD_OP(WR, ATTR, 3),	"resetxattrs")	    \
287 	f(RMXATTR,	__CEPH_OSD_OP(WR, ATTR, 4),	"rmxattr")	    \
288 									    \
289 	/** subop **/							    \
290 	f(PULL,		__CEPH_OSD_OP1(SUB, 1),		"pull")		    \
291 	f(PUSH,		__CEPH_OSD_OP1(SUB, 2),		"push")		    \
292 	f(BALANCEREADS,	__CEPH_OSD_OP1(SUB, 3),		"balance-reads")    \
293 	f(UNBALANCEREADS, __CEPH_OSD_OP1(SUB, 4),	"unbalance-reads")  \
294 	f(SCRUB,	__CEPH_OSD_OP1(SUB, 5),		"scrub")	    \
295 	f(SCRUB_RESERVE, __CEPH_OSD_OP1(SUB, 6),	"scrub-reserve")    \
296 	f(SCRUB_UNRESERVE, __CEPH_OSD_OP1(SUB, 7),	"scrub-unreserve")  \
297 	f(SCRUB_STOP,	__CEPH_OSD_OP1(SUB, 8),		"scrub-stop")	    \
298 	f(SCRUB_MAP,	__CEPH_OSD_OP1(SUB, 9),		"scrub-map")	    \
299 									    \
300 	/** lock **/							    \
301 	f(WRLOCK,	__CEPH_OSD_OP(WR, LOCK, 1),	"wrlock")	    \
302 	f(WRUNLOCK,	__CEPH_OSD_OP(WR, LOCK, 2),	"wrunlock")	    \
303 	f(RDLOCK,	__CEPH_OSD_OP(WR, LOCK, 3),	"rdlock")	    \
304 	f(RDUNLOCK,	__CEPH_OSD_OP(WR, LOCK, 4),	"rdunlock")	    \
305 	f(UPLOCK,	__CEPH_OSD_OP(WR, LOCK, 5),	"uplock")	    \
306 	f(DNLOCK,	__CEPH_OSD_OP(WR, LOCK, 6),	"dnlock")	    \
307 									    \
308 	/** exec **/							    \
309 	/* note: the RD bit here is wrong; see special-case below in helper */ \
310 	f(CALL,		__CEPH_OSD_OP(RD, EXEC, 1),	"call")		    \
311 									    \
312 	/** pg **/							    \
313 	f(PGLS,		__CEPH_OSD_OP(RD, PG, 1),	"pgls")		    \
314 	f(PGLS_FILTER,	__CEPH_OSD_OP(RD, PG, 2),	"pgls-filter")	    \
315 	f(PG_HITSET_LS,	__CEPH_OSD_OP(RD, PG, 3),	"pg-hitset-ls")	    \
316 	f(PG_HITSET_GET, __CEPH_OSD_OP(RD, PG, 4),	"pg-hitset-get")
317 
318 enum {
319 #define GENERATE_ENUM_ENTRY(op, opcode, str)	CEPH_OSD_OP_##op = (opcode),
320 __CEPH_FORALL_OSD_OPS(GENERATE_ENUM_ENTRY)
321 #undef GENERATE_ENUM_ENTRY
322 };
323 
324 static inline int ceph_osd_op_type_lock(int op)
325 {
326 	return (op & CEPH_OSD_OP_TYPE) == CEPH_OSD_OP_TYPE_LOCK;
327 }
328 static inline int ceph_osd_op_type_data(int op)
329 {
330 	return (op & CEPH_OSD_OP_TYPE) == CEPH_OSD_OP_TYPE_DATA;
331 }
332 static inline int ceph_osd_op_type_attr(int op)
333 {
334 	return (op & CEPH_OSD_OP_TYPE) == CEPH_OSD_OP_TYPE_ATTR;
335 }
336 static inline int ceph_osd_op_type_exec(int op)
337 {
338 	return (op & CEPH_OSD_OP_TYPE) == CEPH_OSD_OP_TYPE_EXEC;
339 }
340 static inline int ceph_osd_op_type_pg(int op)
341 {
342 	return (op & CEPH_OSD_OP_TYPE) == CEPH_OSD_OP_TYPE_PG;
343 }
344 static inline int ceph_osd_op_type_multi(int op)
345 {
346 	return (op & CEPH_OSD_OP_TYPE) == CEPH_OSD_OP_TYPE_MULTI;
347 }
348 
349 static inline int ceph_osd_op_mode_subop(int op)
350 {
351 	return (op & CEPH_OSD_OP_MODE) == CEPH_OSD_OP_MODE_SUB;
352 }
353 static inline int ceph_osd_op_mode_read(int op)
354 {
355 	return (op & CEPH_OSD_OP_MODE_RD) &&
356 		op != CEPH_OSD_OP_CALL;
357 }
358 static inline int ceph_osd_op_mode_modify(int op)
359 {
360 	return op & CEPH_OSD_OP_MODE_WR;
361 }
362 
363 /*
364  * note that the following tmap stuff is also defined in the ceph librados.h
365  * any modification here needs to be updated there
366  */
367 #define CEPH_OSD_TMAP_HDR 'h'
368 #define CEPH_OSD_TMAP_SET 's'
369 #define CEPH_OSD_TMAP_CREATE 'c' /* create key */
370 #define CEPH_OSD_TMAP_RM  'r'
371 #define CEPH_OSD_TMAP_RMSLOPPY 'R'
372 
373 extern const char *ceph_osd_op_name(int op);
374 
375 /*
376  * osd op flags
377  *
378  * An op may be READ, WRITE, or READ|WRITE.
379  */
380 enum {
381 	CEPH_OSD_FLAG_ACK =            0x0001,  /* want (or is) "ack" ack */
382 	CEPH_OSD_FLAG_ONNVRAM =        0x0002,  /* want (or is) "onnvram" ack */
383 	CEPH_OSD_FLAG_ONDISK =         0x0004,  /* want (or is) "ondisk" ack */
384 	CEPH_OSD_FLAG_RETRY =          0x0008,  /* resend attempt */
385 	CEPH_OSD_FLAG_READ =           0x0010,  /* op may read */
386 	CEPH_OSD_FLAG_WRITE =          0x0020,  /* op may write */
387 	CEPH_OSD_FLAG_ORDERSNAP =      0x0040,  /* EOLDSNAP if snapc is out of order */
388 	CEPH_OSD_FLAG_PEERSTAT_OLD =   0x0080,  /* DEPRECATED msg includes osd_peer_stat */
389 	CEPH_OSD_FLAG_BALANCE_READS =  0x0100,
390 	CEPH_OSD_FLAG_PARALLELEXEC =   0x0200,  /* execute op in parallel */
391 	CEPH_OSD_FLAG_PGOP =           0x0400,  /* pg op, no object */
392 	CEPH_OSD_FLAG_EXEC =           0x0800,  /* op may exec */
393 	CEPH_OSD_FLAG_EXEC_PUBLIC =    0x1000,  /* DEPRECATED op may exec (public) */
394 	CEPH_OSD_FLAG_LOCALIZE_READS = 0x2000,  /* read from nearby replica, if any */
395 	CEPH_OSD_FLAG_RWORDERED =      0x4000,  /* order wrt concurrent reads */
396 	CEPH_OSD_FLAG_IGNORE_CACHE =   0x8000,  /* ignore cache logic */
397 	CEPH_OSD_FLAG_SKIPRWLOCKS =   0x10000,  /* skip rw locks */
398 	CEPH_OSD_FLAG_IGNORE_OVERLAY = 0x20000, /* ignore pool overlay */
399 	CEPH_OSD_FLAG_FLUSH =         0x40000,  /* this is part of flush */
400 	CEPH_OSD_FLAG_MAP_SNAP_CLONE = 0x80000,  /* map snap direct to clone id */
401 	CEPH_OSD_FLAG_ENFORCE_SNAPC   = 0x100000,  /* use snapc provided even if
402 						      pool uses pool snaps */
403 	CEPH_OSD_FLAG_REDIRECTED   = 0x200000,  /* op has been redirected */
404 	CEPH_OSD_FLAG_KNOWN_REDIR = 0x400000,  /* redirect bit is authoritative */
405 	CEPH_OSD_FLAG_FULL_TRY =    0x800000,  /* try op despite full flag */
406 	CEPH_OSD_FLAG_FULL_FORCE = 0x1000000,  /* force op despite full flag */
407 };
408 
409 enum {
410 	CEPH_OSD_OP_FLAG_EXCL = 1,      /* EXCL object create */
411 	CEPH_OSD_OP_FLAG_FAILOK = 2,    /* continue despite failure */
412 };
413 
414 #define EOLDSNAPC    ERESTART  /* ORDERSNAP flag set; writer has old snapc*/
415 #define EBLACKLISTED ESHUTDOWN /* blacklisted */
416 
417 /* xattr comparison */
418 enum {
419 	CEPH_OSD_CMPXATTR_OP_NOP = 0,
420 	CEPH_OSD_CMPXATTR_OP_EQ  = 1,
421 	CEPH_OSD_CMPXATTR_OP_NE  = 2,
422 	CEPH_OSD_CMPXATTR_OP_GT  = 3,
423 	CEPH_OSD_CMPXATTR_OP_GTE = 4,
424 	CEPH_OSD_CMPXATTR_OP_LT  = 5,
425 	CEPH_OSD_CMPXATTR_OP_LTE = 6
426 };
427 
428 enum {
429 	CEPH_OSD_CMPXATTR_MODE_STRING = 1,
430 	CEPH_OSD_CMPXATTR_MODE_U64    = 2
431 };
432 
433 enum {
434 	CEPH_OSD_WATCH_OP_UNWATCH = 0,
435 	CEPH_OSD_WATCH_OP_LEGACY_WATCH = 1,
436 	/* note: use only ODD ids to prevent pre-giant code from
437 	   interpreting the op as UNWATCH */
438 	CEPH_OSD_WATCH_OP_WATCH = 3,
439 	CEPH_OSD_WATCH_OP_RECONNECT = 5,
440 	CEPH_OSD_WATCH_OP_PING = 7,
441 };
442 
443 const char *ceph_osd_watch_op_name(int o);
444 
445 enum {
446 	CEPH_OSD_BACKOFF_OP_BLOCK = 1,
447 	CEPH_OSD_BACKOFF_OP_ACK_BLOCK = 2,
448 	CEPH_OSD_BACKOFF_OP_UNBLOCK = 3,
449 };
450 
451 /*
452  * an individual object operation.  each may be accompanied by some data
453  * payload
454  */
455 struct ceph_osd_op {
456 	__le16 op;           /* CEPH_OSD_OP_* */
457 	__le32 flags;        /* CEPH_OSD_OP_FLAG_* */
458 	union {
459 		struct {
460 			__le64 offset, length;
461 			__le64 truncate_size;
462 			__le32 truncate_seq;
463 		} __attribute__ ((packed)) extent;
464 		struct {
465 			__le32 name_len;
466 			__le32 value_len;
467 			__u8 cmp_op;       /* CEPH_OSD_CMPXATTR_OP_* */
468 			__u8 cmp_mode;     /* CEPH_OSD_CMPXATTR_MODE_* */
469 		} __attribute__ ((packed)) xattr;
470 		struct {
471 			__u8 class_len;
472 			__u8 method_len;
473 			__u8 argc;
474 			__le32 indata_len;
475 		} __attribute__ ((packed)) cls;
476 		struct {
477 			__le64 cookie, count;
478 		} __attribute__ ((packed)) pgls;
479 	        struct {
480 		        __le64 snapid;
481 	        } __attribute__ ((packed)) snap;
482 		struct {
483 			__le64 cookie;
484 			__le64 ver;     /* no longer used */
485 			__u8 op;	/* CEPH_OSD_WATCH_OP_* */
486 			__le32 gen;     /* registration generation */
487 		} __attribute__ ((packed)) watch;
488 		struct {
489 			__le64 cookie;
490 		} __attribute__ ((packed)) notify;
491 		struct {
492 			__le64 offset, length;
493 			__le64 src_offset;
494 		} __attribute__ ((packed)) clonerange;
495 		struct {
496 			__le64 expected_object_size;
497 			__le64 expected_write_size;
498 		} __attribute__ ((packed)) alloc_hint;
499 	};
500 	__le32 payload_len;
501 } __attribute__ ((packed));
502 
503 
504 #endif
505